Research Article
CPIDM: A Clustering-Based Profound Iterating Deep Learning Model for HSI Segmentation
Table 1
Number of training in addition to the test samples employed for the University of Pavia dataset.
| Class name | Number | Training | Testing | Watershed transform [22] | NN cantered neuro-fuzzy approach [13] | Proposed CPIDM |
| Asphalt | 6631 | 2210 | 4421 | 77.70 | 86.46 | 89.26 | Meadows | 18649 | 6216 | 12433 | 75.30 | 90.17 | 91.49 | Gravel | 2099 | 699 | 1400 | 77.27 | 85.04 | 88.37 | Trees | 3064 | 1021 | 2043 | 92.46 | 96.64 | 96.24 | Painted metal sheets | 1345 | 448 | 897 | 99.63 | 99.78 | 99.81 | Bare soil | 5029 | 1676 | 3353 | 79.50 | 92.39 | 94.89 | Bitumen | 1330 | 443 | 887 | 92.86 | 94.95 | 95.94 | Self-blocking bricks | 3682 | 1227 | 2455 | 76.45 | 85.36 | 90.44 | Shadows | 947 | 14291 | 28521 | 99.62 | 99.65 | 99.89 |
|
|